Wireless communications products have become high volume consumer electronics accessories and are in increasing demand for a broad variety of applications. Features that are increasingly emphasized include reduced power consumption, small form factor, light weight and portability. Many of these products operate in a frequency range extending from about one hundred megahertz to about two gigahertz. As a result, there is a demand to integrate RF receivers and transmitters into high-yield silicon integrated circuit processes to allow a combination of analog, digital and RF functions on a single integrated circuit. "Applications for GaAs and Silicon Integrated Circuits in Next Generation Wireless Communication Systems," by L. M. Burns, IEEE JSSC, Vol. 30, No. 10, October 1995, pp. 1088-1095, discusses examples of monolithic radio receiver and transmitter functions.
Most radio receivers and transmitters require frequency selection components that rely on some form of oscillation to provide frequency selectivity. Devices such as crystals and SAWs that employ mechanical vibration to realize frequency selection also require hermetic packages having interior cavities in order to provide reliable and robust electrical characteristics, particularly in view of environmental insults. Often, the package is more expensive than the component within it. Additionally, multiple packages are required, because the materials useful for these types of frequency selection components do not support active electronic devices, and vice versa. Further, devices that rely on mechanical oscillation also use materials having different packaging requirements than do active electronic devices.
Electronic circuits that do not rely on mechanical vibrations for frequency selection characteristics often rely instead on electrical resonances to provide frequency selectivity. Practical electrical resonators in this frequency range require a combination of capacitance and inductance. Of these, inductance is particularly difficult to realize in compact form together with reasonably high quality factor, or "Q." Q is often defined as the amount of energy stored divided by the amount of energy dissipated per cycle, but can also be defined as a center frequency divided by a three dB bandwidth of a frequency response. The latter definition is used herein in instances where the former is inapplicable.
Known approaches for realizing monolithic inductance include spiral inductors, transmission lines and bond wires. For example, "A 1.8 GHz Low-Phase-Noise Spiral-LC CMOS VCO," by J. Cranickx and M. Steyaert, 1996 Symp. on VLSI Cir. Dig. Tech. Papers, pp. 30-31 describes a spiral inductor approach that achieves a Q of 5.7 near two gigahertz. "Integrated Passive Components in MCM-Si Technology and their Applications in RF-Systems," by J. Hartung, 1998 Int. Conf. on Multichip Modules and High Density Packaging, IEEE Cat. No. 0-7803-4850-8/98, pp. 256-261, reports Qs and their frequency dependence for spiral inductors vs. substrate resistance, with highest Qs and self-resonant frequencies for spiral inductors fabricated on higher-resistivity substrates. A recent overview of spiral inductive components, entitled "Analysis, Design, and Optimization of Spiral Inductors and Transformers for Si RF IC's," by A. Niknejad and R. Meyer, IEEE JSSC, Vol. 33, No. 10, October 1998, pp. 1470-1481, gives examples of Qs having peak values around five and inductances of up to about ten nanoHenrys for spiral inductors fabricated on silicon.
Transmission line approaches to realizing monolithic inductance tend to be bulky and relatively lossy in this frequency range. Bond wires can provide Qs ranging from 11 to 15, as described in "A 1V, 1.8 GHz, Balanced Voltage-Controlled Oscillator with an Integrated Resonator," by D. A. Hitko et al., Proc. Symp. Low Power Electr. and Des., pp. 46-51 (1997). Bond wire inductors tend to be relatively large compared to other integrated circuit components, but do permit the surface area beneath them to be used to fabricate other integrated circuit elements prior to bond wire installation. Bond wire inductors also require bond pads, which are relatively large, and which preclude use of their area for other purposes. None of these approaches provide the combination of small form factor, high Q and packageability needed for many applications.
Another approach to providing a frequency selection function in monolithic form relies on impedance transformations that are possible with active circuits, i.e., circuits including transistors. U.S. Pat. No. 5,175,513, entitled "Oscillator Circuit Employing Inductive Circuit Formed of Field Effect Transistors" and issued to S. Hara describes an example using MESFETs. U.S. Pat. No. 5,726,613, entitled "Active Inductor" and issued to H. Hayashi et al. and "A Novel Broad-Band MMIC VCO Using an Active Inductor," by H. Hayashi and M. Maraguchi, IEICE Trans. Fundamentals, Vol. E81-A, No. 2, February 1998, pp. 224-229, describe similar approaches. While these approaches do provide compact circuits, they use GaAs MESFETs, which are not as manufacturable as CMOS FETs and which are not cost-competitive with silicon integrated circuits. Additionally, it is much more expensive to provide complex ancillary functions on GaAs substrates, such as may be realized using digital circuitry, than is the case with silicon substrates.
Some oscillators do not require either inductance to form an electrical resonance or mechanical oscillations to achieve frequency selectivity or control. "A Study of Phase Noise in CMOS Oscillators" by B. Razavi, IEEE JSSC, Vol. 31, No. 3, March 1996, pp. 331-343, compares several different types of CMOS-compatible oscillators. A first type is a ring oscillator employing a number of inverter stages coupled in a ring configuration. Ring oscillators are used in some systems, however, ring oscillators have limited voltage tuning capability, require a relatively large number of FETs and do not provide a sinusoidal output signal. A second type of oscillator that is discussed is a relaxation oscillator. The relaxation oscillator is simpler than the ring oscillator and requires less area to implement in an integrated circuit, but also has limited tunability.
Therefore, there is a need for monolithic circuitry that provides voltage-controlled oscillators and that is compatible with cost-effective approaches to providing other circuit functions.